Short-term Forecasting for Charging Pile and Photovoltaic Loads Based on LSTM-Seq2seq Model in Extreme Weather

In order to solve the problem of low prediction accuracy for charging pile (CP) and photovoltaic (PV) load in extreme weathers, a short-term load forecasting model based on the Pearson correlation analysis method and LSTM-Seq2seq network is proposed in this paper. Firstly, meteorological data sets and corresponding load data under extreme weather conditions were filtered using specific meteorological indicators. Next, the meteorological indicators were identified using Pearson correlation analysis, which show a high correlation with the load were selected. Then, the LSTM-Seq2Seq model is used to capture the temporal relationships between the load and the features. Finally, the above features were input into the model to obtain improved short-term load forecasting results. The effectiveness of the model is verified by case studies.